evaluation of multiple-channel ofdm based...
TRANSCRIPT
Evaluation of multiple-channel OFDM based airborne ultrasoniccommunications
Wentao Jiang∗, William M. D. Wright∗
School of Engineering - Electrical and Electronic Engineering, University College Cork, Cork, Ireland
Abstract
Orthogonal frequency division multiplexing (OFDM) modulation has been extensively used in both wired
and wireless communication systems. The use of OFDM technology allows very high spectral efficiency data
transmission without using complex equalizers to correct the effect of a frequency-selective channel. This work
investigated OFDM methods in an airborne ultrasonic communication system, using commercially available
capacitive ultrasonic transducers operating at 50 kHz to transmit information through the air. Conventional
modulation schemes such as binary phase shift keying (BPSK) and quadrature amplitude modulation (QAM)
were used to modulate sub-carrier signals, and the performances were evaluated in an indoor laboratory en-
vironment. Line-of-sight (LOS) transmission range up to 11 m with no measurable errors was achieved using
BPSK at a data rate of 45 kb/s and a spectral efficiency of 1 b/s/Hz. By implementing a higher order mod-
ulation scheme (16-QAM), the system data transfer rate was increased to 180 kb/s with a spectral efficiency
of 4 b/s/Hz at attainable transmission distances up to 6 m. Diffraction effects were incorporated into a model
of the ultrasonic channel that also accounted for beam spread and attenuation in air. The simulations were
a good match to the measured signals and non-LOS signals could be demodulated successfully. The effects of
multipath interference were also studied in this work. By adding cyclic prefix (CP) to the OFDM symbols, the
bit error rate (BER) performance was significantly improved in a multipath environment.
Keywords:
Airborne ultrasonic communication, binary phase shift keying (BPSK) and quadrature amplitude modulation
(QAM) modulation, capacitive ultrasonic transducer, multipath interference, orthogonal frequency division
multiplexing (OFDM)
1. Introduction
The use of modulated ultrasound as a way of communication has existed for many years. Most ultrasonic
communication systems have been developed for underwater applications because the physical properties of
liquids tend to allow sound waves travelling via molecular vibrations to cover relatively long distances. However,
there is limited information published in the area of ultrasonic communications in air. This is because sound
propagation in air is more difficult, in that it is much more susceptible to changes in atmospheric conditions and
attenuation is significant. As ultrasonic signals do not travel long distances in air, it is not surprising that most
ultrasonic communication systems are designed for indoor use [1, 2, 3, 4]. One particular aspect of interest is the
high security features that are inherent as ultrasound is virtually impossible to detect from outside the room,
∗Corresponding authorEmail addresses: [email protected] (Wentao Jiang), [email protected] (William M. D. Wright)
Preprint submitted to Elsevier May 21, 2016
eliminating the risks of wireless interception such as Bluesniping and jamming attacks [5, 6]. In addition, in
areas with electronics sensitive to electromagnetic interference, ultrasonic transmission is a desirable alternative
compared to the use of radio frequency methods.
The attenuation of ultrasound in air at high frequencies is high, which limits the potential transmission
bandwidth and range of the communication system. Therefore, to achieve reasonable data rates and practical
transmission ranges, it is necessary to apply a suitable communication technique which can make full use of
the limited available bandwidth of ultrasonic transducers that operate in air. Orthogonal frequency division
multiplexing (OFDM) is a highly efficient form of multicarrier modulation which is extensively used in modern
telecommunications, ranging from asymmetric digital subscriber line (ADSL) modem technology to 802.11 Wi-
Fi wireless systems [7]. By spreading the data across orthogonal frequency bands, OFDM is also an effective
solution to inter-symbol interference (ISI) caused by a dispersive channel. Quadrature amplitude modulation
(QAM) is most often used in OFDM to further improve the bandwidth efficiency and data transfer rate. In
this paper, both binary phase shift keying (BPSK) and higher order modulation schemes (4-QAM and 16-
QAM) are evaluated for ultrasonic communications in air. Problems that occur in OFDM transmissions are
also investigated and solutions are given. Section II provides a review of the relevant academic literature in the
area of airborne ultrasonic communications and the state of the art of this work in the field. A typical OFDM
system for wireless applications is presented in Section III. This is followed by the apparatus and experimental
results in Section IV and Section V, respectively. Section VI gives the conclusions of the work.
2. Related work
The first generation batteryless, wireless remote control system for TV receivers was designed by Robert
Adler for Zenith televisions in the 1950s [1]. Distinctive high frequency sounds at ultrasonic frequencies of
around 40 kHz were produced by striking different lengths of aluminium rod with a set of hammer buttons.
These sounds were decoded at the TV, which changed channels appropriately. In [2], an indoor positioning
system was developed by emitting frequency shift keying (FSK) modulated ultrasonic signals. In this work,
the attainable transmission range achieved was up to 20 m, with a data rate of 100 b/s on a 4 kHz bandwidth
at 40 kHz. Another indoor positioning system achieved a ranging accuracy of 2 cm using piezo film ultrasonic
transducers [3]. BPSK modulated direct sequence code division multiple access (CDMA) signal structure was
used in this work, and achieved a data rate of 20 kb/s on a 76 kHz bandwidth at 50 kHz. Later work on
multichannel data communication investigated high frequency amplitude shift keying (ASK) and on-off keying
(OOK) at 250 kHz with a 0.5 m transmission distance and achieved an 80 kb/s data transfer rate using
prototype transducers [8]. In [4], the authors studied the performance of air-coupled ultrasonic transmission
over directional short-range links using OOK, binary frequency shift keying (BFSK) and BPSK modulations.
The system started to experience error decoding at a range of 1.6 m and the maximum data rate achieved
was 83 kb/s using a physical hard-wired synchronization link between the transmitter and the receiver in a
laboratory environment. Another later study by the same authors implemented single-channel quadrature
phase shift keying (QPSK) modulation using the same custom-made capacitive transducers in [4] and achieved
a bit rate of 200 kb/s over 1.2 m in air in a line-of-sight (LOS) manner [5]. Recent work looked at a 2-channel
OFDM scheme for use in localization systems [9]. However, as piezoelectric transducers were used, the achieved
data rate was only 5.7 kb/s and the measured packet error rate was 13% at a distance of 18 m.
2
In this work, multiple-channel OFDM with different order baseband modulation schemes is investigated
to achieve the longest error-free decoding range and highest data rate possible using a pair of commercially
available capacitive ultrasonic transducers at 50 kHz. The essential problems due to phase noise and multipath
interference that OFDM may encounter in practical use are analysed and appropriate solutions are given. Of
particular interest are the effects of multipath interference and diffracted non-LOS performance.
3. OFDM system description
3.1. OFDM modulation and demodulation
To enable the largest amount of data to be communicated using a finite frequency range, conventional
frequency division multiplexing (FDM) modulation uses relatively small symbol durations. The system then
becomes more susceptible to loss of information from impulse noise, multipath distortion and other impairments.
FDM also allows each channel to be band-limited to a specific frequency range by implementing a pulse-
shaping filter [10]. However, this technique introduces unnecessary waste of the frequency spectrum in guard
spaces between the sub-channels. It also increases the complexity of each sub-carrier modulator. OFDM
modulation decreases the rate of the original data stream in all parallel sub-carrier channels in order to mitigate
multipath interference. By generating overlapping but non-interfering sub-channels, it also maximises the
spectral efficiency. In a typical OFDM system sending complex data {dn}, N harmonic sinusoids that are
orthogonal on a period Ts of the fundamental sub-carrier signal are modulated and combined in parallel, giving
the complex envelope s(t) of the OFDM signal:
s(t) =
N−1∑n=0
dnej2πfntΠ(t), (1)
Π(t) =
1, 0 < t ≤ Ts
0, otherwise.
(2)
Here, fn = n/Ts, and Π(t) represents a rectangular pulse with a duration of Ts. The rectangular window has
a sinc-shape spectrum which makes a sub-channel spectrum heavily overlap its neighbours. Take N samples of
the pulse s(t) within Ts, the OFDM signal can then be obtained efficiently by an inverse fast Fourier transform
(IFFT) [11] of the data sequence {dn}:
s(m) =
N−1∑n=0
dnej 2πN nm for m ∈ [0, N − 1]
= N × IFFT{dn}.
(3)
At the receiver, the demodulation for an OFDM signal is essentially the reverse process of the modulation. The
fast Fourier transform (FFT) is therefore performed on the sampled OFDM signal to recover the data sequence
{dn}, avoiding the need for complex sub-channel filters.
3.2. Synchronization
Synchronization is a fundamental requirement for any wireless communication system to work properly. Once
a data packet is detected, the receiver performs a synchronization operation that identifies the starting point of
3
the packet. Therefore, a preamble is needed at the start of the data packet for wireless synchronization. A linear
frequency modulation (LFM) signal whose frequency sweeps linearly from low to high within a certain time is
used in this work. It is also the most commonly used signal in practical radar systems for range detection [12].
By performing correlation of a known LFM signal and the received signal using a matched-filter at the receiver,
the maximum energy of the matched-filter output then indicates the start of the arriving data sequence. The
main advantage of using a LFM signal is that it provides precise timing resolution for frame synchronisation as
the matched filter maximises the signal-to-noise ratio (SNR) of the signal [10, 13].
3.3. Phase noise
One of the non-ideal effects that can occur in a real OFDM system is symbol timing offset (STO) which
results in a rotated phase in the received constellation [14]. A known LFM preamble is located at the beginning
of the data sequence to detect the boundary of the OFDM symbol after matched filtering. However, inaccuracy
may still remain, and this leads to an offset in the FFT window location. Assuming the STO is δ samples, the
received signal will then be a shifted version of s(m) in (3). After applying the FFT, the recovered complex
number sequence yn for each OFDM symbol can be derived as:
yn =
N−1∑m=0
{s(m+ δ)} · e−j 2πN mn
=
N−1∑m=0
{N−1∑n=0
dnej 2πN n(m+δ)
}· e−j 2π
N mn
= dn · ej2πN nδ, n ∈ [0, N − 1].
(4)
As (4) shows, an STO of δ samples in the time domain incurs a phase offset of 2πN nδ in the frequency domain,
which is proportional to the sub-carrier index n, as well as the STO. This means that the symbol timing error
does not destroy the orthogonality of the sub-carriers, and the effect of the timing error is a phase rotation
which changes linearly with the carrier index.
STO can be corrected by applying the maximum likelihood (ML) approach [10] to the data-aided phase
recovery scheme. Mathematically, the maximisation can be expressed as:
δ̂ = arg maxδ
{N−1∑m=0
s(m)s(m+ δ)
}. (5)
The sum in equation (5) is actually the time-correlation between the received signal s(m + δ) and the signal
of interest s(m). Therefore, in this particular case, the ML approach is equivalent to the maximisation of the
correlation.
3.4. OFDM in a multipath environment
The main advantage of OFDM over other forms of modulation is that it is much easier to make it immune
to ISI caused by multipath propagation. In a linear dispersive channel, the received signal sequence, r(m), can
be expressed as the sum of several versions of the transmitted signal with different gains and delays:
r(m) =
L−1∑i=0
αis(m− δi), (6)
4
where L is the number of multipath components, αi are path attenuation factors and δi are path delays. ISI could
be easily eliminated by adding guard intervals (GI) between different OFDM symbols. As long as the length of
the GI is larger than the maximum estimated delay spread, the effect of ISI can then be completely removed.
However, if the GI is left empty, it will lead to another type of interference called inter-carrier interference (ICI)
as the zero-padded waveform would destroy the orthogonality between sub-carriers. Therefore, to prevent both
ISI and ICI, an OFDM symbol is usually cyclically extended into the GI. This is achieved by taking the last
few samples at the end of each symbol, and adding these samples to the start of the symbol. This cyclic prefix
(CP) preserves the sub-carrier orthogonality at the expense of wasting some energy, as the power transmitted
in the added symbols is not used at the receiver. At the same time, the system data rate is also reduced due to
the extension of the total OFDM packet length.
To analyse the effects of a multipath channel on an OFDM signal, a simplified model was developed ignoring
GI and background noise. Considering only the direct path signal and one reflected signal with a time delay δ
and an attenuation factor of α, the following signal r(m) is detected by the receiver:
r(m) = s(m) + αs(m− δ)
=
N−1∑n=0
dnej 2πN nm + α
N−1∑n=0
dnej 2πN n(m−δ)
=
N−1∑n=0
dn(1 + αe−j2πN nδ)ej
2πN nm, m ∈ [0, N − 1].
(7)
Applying demodulation by the FFT, the transmitted data yn can be recovered from the received signal r(m):
yn = FFT{r(m)}
= dn(1 + αe−j2πN nδ)
= dnH(n),
(8)
where H(n) = 1 + αe−j2πN nδ for n ∈ [0, N − 1]. As equation (8) indicates, the delayed signal only changes the
amplitude |H(n)| and the phase arg{H(n)} of each sub-carrier. If the path delay δ is smaller than the length
of the GI, no ICI is introduced, and the orthogonality between sub-carriers still holds. Assuming that a 4-QAM
modulated signal is being transmitted through this channel, the constellations with different delay spread and
attenuation of the detected signal are shown in Fig. 1. As can be seen, the effect of the multipath channel is to
expand the constellations on each sub-carrier into different circles surrounding the target points. The number
of points on each circle depends on the values of δ and N , and the radius of each circle depends on the value of
α. If α is small, the transmitted data can still be correctly decoded. If α is large, and the circles cross over each
other, an equaliser is then required to correct the channel distortion. According to equation (8), each received
sub-carrier experiences a complex gain H(n) due to the channel. In order to undo this effect, the transmitted
data can be recovered from the received signal by multiplying dn by 1/H(n) which is just a single complex
multiplication in the frequency domain. It greatly simplifies the equalisation process, especially compared to
the adaptive filtering used in single carrier demodulations.
5
−1 0 1−1.5
−1
−0.5
0
0.5
1
1.5
In−Phase
Qua
drat
ure
N=1024,NGI
=256,δ=32,α=0.3
−1 0 1−1.5
−1
−0.5
0
0.5
1
1.5
In−Phase
Qua
drat
ure
N=1024,NGI
=256,δ=32,α=0.6
−1 0 1−1.5
−1
−0.5
0
0.5
1
1.5
In−Phase
Qua
drat
ure
N=1024,NGI
=256,δ=128,α=0.3
−1 0 1−1.5
−1
−0.5
0
0.5
1
1.5
In−Phase
Qua
drat
ure
N=1024,NGI
=256,δ=128,α=0.6
Figure 1: Spread of the 4-QAM OFDM constellations with different delay and amplitude of the reflected signal.
4. Apparatus and experiment set-up
The OFDM method previously described was evaluated using a pair of commercially available SensComp
series 600 capacitive ultrasonic transducers [15] for wireless communication. The transducers are composed of a
metallised Kapton membrane and a rigid contoured metal backplate with an aperture size of 38.4 mm. They are
specially designed for operation in air at an ultrasonic frequency of 50 kHz with a beam angle of 15◦ at -6 dB.
Fig. 2 illustrates the experimental arrangement for ultrasonic wireless communications in air. The data signal
was modulated by MATLAB (MathWorks) running on a PC before sending to a TTi TGA 12102 arbitrary
waveform generator using GPIB. The signal was then amplified using a Falco WMA-300 high voltage amplifier,
and combined with a superimposed +200 V DC bias voltage generated by a Delta Elektronika SM3004-D Power
supply. The receiver, identical to the transmitter, was connected to a Cooknell CA6/C charge amplifier powered
by a Cooknell SU2/C power supply unit, and followed by a high performance PicoScope 6403A PC oscilloscope.
After that, the captured signal was sent back to another PC via a USB interface for signal processing.
TTi TGA12102
100MHz AWG
Delta Elektronika SM3004-D
Power Supply
Cooknell CA6/C
Charge Amplifier
PicoScope 6403A PC
Oscilloscope
Falco High
Voltage
Amplifier
WMA-300
Transmitter Receiver
Air Gap
GPIB USB
Cooknell SU2/C
Power Supply Unit
DATA
SYNC.
Figure 2: Schematic diagram of the experimental set-up.
The overall system amplitude and phase response through an air gap of 2 m are illustrated in Fig. 3 (a) and
(b) respectively. The received impulse was measured by sending a pulse generated by a Panametrics 500PR
6
Table 1: Parameters of the OFDM signal used in the experiment when transducers are aligned
Parameters Settings
Sampling rate fs = 10 MHz
Centre frequency fc = 77 kHz
Signal bandwidth B = 45 kHz
OFDM symbol duration Ts = 1 ms
Sub-carrier spacing ∆f = 1 kHz
Number of sub-carriers N = 45
pulser from the transmitter to the receiver. As can be seen in Fig. 3 (a), the prominent spectra ranges from
50 kHz up to about 110 kHz. Fig. 3 (b) shows that the phase response of the channel is nearly linear at
frequencies between 30 and 170 kHz. The recorded received signal amplitude was typically around 46.6 mV rms
at 2 m, and the rms amplitude of the background noise was at about 1.2 mV.
0 50 100 150 200−50
−40
−30
−20
−10
0
Frequency(kHz)
Mag
nitu
de(d
B)
0 50 100 150 200−400
−300
−200
−100
0
Frequency(kHz)
Pha
se(r
adia
ns)
(a)
(b)
Figure 3: System characteristics: (a) system response and (b) phase response.
All experiments were conducted in an indoor laboratory environment with negligible air turbulence to
influence the data transmission. Transmitter and receiver transducers were separated in a LOS manner, and
the transmission bandwidth used was from 55 to 99 kHz for best results on a trial and error basis. Parameters of
the OFDM signal are listed in Table 1. Sub-carrier signals were modulated using BPSK and two levels of QAM
(4-QAM and 16-QAM). The three different modulation schemes from low order to high order were evaluated
and compared in the experiments. Note that both 4-QAM and 16-QAM were Gray coded so that for most
symbol errors there is only one bit received in error [16].
5. Results and discussion
5.1. Line-of-sight (LOS) transmission
Initial transducer separation was set at 1 m, and different modulated signals with SNRs incremented in 1 dB
steps from 5 dB to 18 dB were transmitted. Due to hardware restriction, 90 packets of 45-subchannel OFDM
7
Table 2: Performances of BPSK, 4-QAM and 16-QAM modulations in an OFDM system
BPSK 4-QAM 16-QAM
Error-free decoding range (m) 11 9 6
Data rate (kb/s) 45 90 180
Bandwidth efficiency (b/s/Hz) 1 2 4
signal were generated and transmitted continuously through the air channel. The process was repeated 10 times,
and then averaged. Therefore, 162000 transmitted bits were observed, allowing the minimal measurable BER
of 6.17 × 10−6. Fig. 4 compares the BER performances of the three modulation schemes in both theory and
experiment. As expected, when the noise becomes smaller compared to the signal level, less error occurs. Higher
order modulation schemes require a larger SNR to transmit data with fewer bit errors. However, large gaps
exist between the theoretical and the experimental BER curves. Note that the theoretical curves are the best
achieved BER performance for all three modulation schemes. The discrepancy between theory and experiment
results is due to a few reasons. One major reason is the hardware implementation penalty as the received
signal was undersampled by 2.34% at the oscilloscope. Therefore, inaccurate timing occurs, and it reduces the
orthogonality of the OFDM signal. The second reason is the ISI introduced by multipath interference. Another
reason is the bandwidth limitation of the selected ultrasonic transducers. Wider bandwidth transducers may
improve the BER performance, especially for high-order modulation schemes. OFDM packet errors for the three
modulation schemes were also measured with increased transmission ranges. Error-free decoding was achieved
at different ranges for different modulations as listed in Table 2. As the noise margin of BPSK is larger than
either 4-QAM or 16-QAM, BPSK holds a more reliable transmission link which is up to 11 m with no measurable
bit errors, followed by the error-free decoding ranges for 4-QAM and 16-QAM of 9 m and 6 m, respectively.
Table 2 also presents the maximum data transfer rates and the corresponding bandwidth efficiencies for the
three modulation schemes. The advantage of using QAM compare to BPSK is that it is able to carry more
bits of information per symbol. By increasing the order format of QAM, the data rate of the system as well
as the bandwidth efficiency can be significantly increased. It also indicates that each individual sub-carrier can
be treated independently from the others in terms of the modulation method used in a self-adaptive system.
Thus, a sub-carrier with more channel attenuation can use a more resilient modulation like BPSK or 4-QAM
while sub-carriers with better channels can use a higher order modulation like 16-QAM.
To visualise the conditions of the received OFDM signals, constellation diagrams of BPSK, 4-QAM and
16-QAM before and after the phase correction and channel equalisation are given in Fig. 5. As can be seen from
Fig. 5 (a) - (c), the raw received signal constellations are corrupted as the constellation points are rotated and
spread across the detection boundaries due to phase noise and the effect of the ultrasonic channel. After phase
noise correction and channel equalisation, as shown in Fig. 5 (d) - (f), both the phase rotation and amplitude
distortion have been corrected with the aid of a known pilot signal. Note this pilot signal contains randomly
generated bits modulated in the same way as the OFDM signal, and was transmitted along with the information
signal.
8
6 8 10 12 14 16 1810
−8
10−7
10−6
10−5
10−4
10−3
10−2
10−1
100
SNR (dB)
BE
R
Theor. BPSKTheor. QPSKTheor. 16QAMExper. BPSKExper. QPSKExper. 16QAM
Figure 4: Theoretical and experimental bit error rate performance of three different OFDM modulation schemes at 1 m range.
5.2. Reflected and diffracted signals
The maximum beam divergence angle of the ultrasonic signal at 55 kHz used in the experiment was 11.4◦.
Thus, the emitted ultrasonic signals were relatively directional. Similarly, the receiver can only detect signals
with small incidence angles. In this case, when the transducers are aligned in a LOS manner, most undesired
reflections can be eliminated. In the situations when the transducer centre normals are not coincident and
the receiver transducer is close to possible reflectors such as the floor and walls, reflected signals can then be
detected. Diffraction also occurs in practice. This can be seen as the deflection or ‘bending’ of the waves, caused
by an obstacle in the propagation path or by nonhomogeneity of the air [17]. The effects of diffraction can be
evaluated by measuring signal levels for different frequencies diffracted from a sharp-edged screen. Fig. 6 shows
the configuration of transmitter transducer (Tx), receiver transducer (Rx), and a blocking screen, where the
lateral displacement u of the receiver can vary with the transmitter and receiver transducer faces in parallel
(position A). It should be noted that the screen occluded half of the transmitted aperture when u = 0, and
only diffracted signals could be obtained when u ≥ 0.0192 m and no LOS signal could be received. Fig. 7 shows
the resulting signal attenuation at the receiver Rx as a function of u, for the highest signal frequency (99 kHz)
and the lowest signal frequency (55 kHz) used in the experiment. It can be seen that for both low and high
frequency signals, the larger the lateral displacement, the smaller amount of diffracted energy that is received,
as expected. In addition, the low frequency signals are diffracted more than the high frequency signals and can
be detected at larger values of lateral displacement. However, when the lateral displacement is as small as 0.2 m
for example, the signal attenuation for high and low frequency waves was approximately -30 dB and -19 dB
respectively.
The reduction of sound pressure level due to diffraction around the edge of a barrier can be described in
terms of a Fresnel number [17, 18], which is given by
N =2
λ(d1 + d2 − d) (9)
where λ is the wavelength of sound and lengths d1, d2, and d are as shown in Fig. 6. The attenuation by
diffraction Ad, in dB, is then approximated as a function of the Fresnel number
9
−2 0 2−2
−1
0
1
2
In−Phase(a)
Qua
drat
ure
−2 0 2−2
−1
0
1
2
In−Phase(d)
Qua
drat
ure
−2 0 2
−2
−1
0
1
2
In−Phase(b)
Qua
drat
ure
−2 0 2
−2
−1
0
1
2
In−Phase(e)
Qua
drat
ure
−5 0 5−5
0
5
In−Phase(c)
Qua
drat
ure
−5 0 5−5
0
5
In−Phase(f)
Qua
drat
ure
Figure 5: Received constellations before (a) - (c) and after (d) - (f) phase noise correction and channel equalisation for BPSK-OFDM,
4QAM-OFDM and 16QAM-OFDM at 11 m, 9 m and 6 m respectively.
Ad = 10log10(20N). (10)
These diffraction effects were incorporated into an earlier model of the ultrasonic channel [19] that accurately
predicted the LOS channel response in air, accounting for beam spread and attenuation along the propagation
path. Equation (10) yields the two attenuation curves for 55 kHz and 99 kHz signals in Fig. 7. As can be seen,
the theoretical prediction matches the experimental data well. Note that the model only considers a point source
and receiver, therefore the actual received signal energy may decrease faster than the theoretical prediction with
the increase of lateral displacement as the effective reception area of the receiver is an ellipse rather than a circle.
By using the same diffraction setup in Fig. 6, the BER results for all three modulation schemes with increasing
lateral displacements up to 0.2 m are shown in Fig. 8. As can be seen, the BER curves for all three modulations
ascend rapidly with lateral displacement. 16-QAM, with the most spreading of the spots in the constellation and
hence with reduced noise immunity, experienced the most error decoding at all ranges. Lower-order modulation
schemes, 4-QAM and BPSK, proved their superiority over 16-QAM when signals are diffracted around a sharp-
edged screen, with first bit errors occurring at 0.1 m and 0.125 m, respectively. When the lateral displacement
of the receiver increased to 0.2 m, the BERs for all three modulations reached a level close to 30%. It indicates
that BPSK and 4-QAM modulation schemes work better using only diffracted signals than 16-QAM in terms of
BER performance. However, non-LOS transmission with no measurable errors was achieved with BPSK-OFDM
and 4QAM-OFDM.
One situation that occurs when the face of the receiver is angled towards the edge of the screen barrier
is illustrated in Fig. 6 position B. The theoretical signal attenuation for 55 kHz and 99 kHz signals with an
increasing oblique angle θ from 0◦ to 40◦ in a 5◦ step are compared with experimental results in Fig. 9, and
good agreement was found between the two. Accordingly, the BER results for the three modulation schemes are
illustrated in Fig. 10. It shows a similar overall trend as the curves in Fig. 8. For 4-QAM and BPSK, the two
10
modulation methods experienced signal error-free decoding with oblique angles up to 10◦ and 20◦, respectively.
Rx
Tx
1 m 1 m
Screen
Rx
d
d1
d2u
E
A
B
Figure 6: Configuration of transmitter Tx and receiver Rx where ultrasonic waves diffract around the edge of a blocking screen.
Position A: transducer faces parallel, position B: receiver rotates and faces the edge E.
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
−30
−25
−20
−15
−10
−5
0
Lateral displacement(m)
Atte
nuat
ion(
dB)
Received 55kHz signalReceived 99kHz signal55kHz theoretical prediction99kHz theoretical prediction
Figure 7: Theoretical and experimental signal attenuation for 55 kHz and 99 kHz ultrasonic waves with varied lateral displacement
in a diffraction setup consisting of a sharp-edged screen.
5.3. Transmission in a multipath environment
In practice, the effect of multipath reflections cannot be ignored. It can lead to the superposition of several
time-delayed and attenuated copies of the transmitted signal at the receiver. As a result, it may introduce
spreading of the received signal in time and constructive and destructive interference at various frequencies,
leading to frequency selectivity. A typical reflection pattern in the time domain is shown in Fig. 11 when a
Hamming-windowed 10-cycle 55 kHz sinusoid signal was transmitted through the system in the presence of a
reflecting surface. The enlarged region in Fig. 11 shows two strong reflections followed by additional signals with
much smaller amplitudes within 2 ms of the direct signal. Each pattern of peaks corresponded to the arrival of
acoustic reflections at the receiver at different moments in time. It should be noted that all the reflections are
relatively weak compared with the direct signal due to the directivity of the receiver. To combat both ISI and
ICI, a CP with a length of 2 ms was then added to the OFDM symbols. Accordingly, the length of the OFDM
symbols should also be extended. Details of the parameter settings are listed in Table 3. Since the OFDM
symbol length was extended to 8 ms, to preserve sub-carrier orthogonality, the sub-channel width was reduced
11
0 0.05 0.1 0.15 0.210
−5
10−4
10−3
10−2
10−1
100
Lateral displacement (m)
Bit
erro
r ra
te
BPSK OFDM4−QAM OFDM16−QAM OFDM
error free below 0.125 m
error free below 0.1 m
Figure 8: BERs with varied lateral displacement in a diffraction setup consisting of a sharp-edged screen using BPSK, 4-QAM and
16-QAM modulations, showing error-free non-LOS transmission ranges for BPSK-OFDM and 4QAM-OFDM.
0 5 10 15 20 25 30 35 40
−30
−25
−20
−15
−10
−5
0
Oblique angle (degree)
Atte
nuat
ion
(dB
)
Received 55kHz signalReceived 99kHz signal55kHz theoretical prediction99kHz theoretical prediction
Figure 9: Theoretical and experimental signal attenuation for 55 kHz and 99 kHz ultrasonic waves with varied oblique angles in a
diffraction setup consisting of a sharp-edged screen.
to only 125 Hz with a total number of 360 sub-carriers in the range of 55 to 99 kHz. The tested error-free
decoding ranges for BPSK, 4-QAM and 16-QAM were maintained the same as listed in Table 2 using this new
channel arrangement. However, after adding the CP, total data transfer rates for all three modulation schemes
were reduced by 20%. To evaluate the performance of using CP in a multipath environment, BERs of the
OFDM signals with CP and without CP are compared in Fig. 12. It is observed that there is an improvement
in BER after adding CP, especially when the received signal has a large SNR. Note that only the performance
of 16-QAM OFDM signals are presented here as an illustrative example. Accordingly, the constellations of the
equalised 16-QAM signals without and with CP at a SNR value of 32 dB are shown in Fig. 13 (a) and (b),
respectively. As can be seen, more concentrated constellation points with less spreading were extracted from
the OFDM signal after adding CP.
12
0 5 10 15 20 25 30 35 4010
−4
10−3
10−2
10−1
100
Oblique angle (degree)
Bit
erro
r ra
te
BPSK OFDM4−QAM OFDM16−QAM OFDM
error free below 10°
error free below 20°
Figure 10: BERs with varied oblique angles when the receiver is facing towards the edge of the blocking screen using BPSK, 4-QAM
and 16-QAM modulations.
5.5 6 6.5 7 7.5 8
−200
−150
−100
−50
0
50
100
150
200
Time(ms)
Am
plitu
de(m
V)
6.5 7 7.5 8
−10
−5
0
5
10
Figure 11: Received direct signal and reflected signals in a multipath environment.
6. Conclusions
This paper evaluated an airborne multiple-channel OFDM data communication system using a pair of
commercially available capacitive ultrasonic transducers. The main advantage of the multiple-channel OFDM
modulation scheme was that it enabled channel equalization in the frequency domain, thus eliminating the need
for potentially complex time-domain equalizers. It also provided a data transmission robust against both ICI
and ISI caused by the multipath channel. Due to hardware limitations, most air-coupled capacitive ultrasonic
transducers have a relatively narrow bandwidth to provide data communication with an acceptable transfer rate.
OFDM modulation can maximise bandwidth utility by allowing sub-carriers to overlap each other. Different
order baseband modulation schemes from BPSK, 4-QAM and 16-QAM were implemented in the OFDM system
using an ultrasonic band from 55 to 99 kHz and individual performances were evaluated. The maximum data
rate achieved in this work was 180 kb/s using 16-QAM OFDM in a LOS manner. BPSK-OFDM experienced the
lowest decoding errors when using the same signal power, thus the longest error-free decoding range achieved
was up to 11 m with a data rate of 45 kb/s. It was shown by experiment that all three modulation schemes
investigated could operate in a non-LOS configuration after diffracting around a sharp-edged screen, and that
13
Table 3: Parameters of the OFDM signal used in the experiment in a multipath environment
Parameters Settings
Sampling rate fs = 10 MHz
Centre frequency fc = 77 kHz
Signal bandwidth B = 45 kHz
OFDM symbol duration Ts = 8 ms
Cyclic prefix length Tcp = 2 ms
Sub-carrier spacing ∆f = 125 Hz
Number of sub-carriers N = 360
22 24 26 28 30 32
10−4
10−3
10−2
10−1
SNR(dB)
BE
R
Without CPWith CP
Figure 12: Bit error rate performance for 16-QAM OFDM modulation with and without cyclic prefix..
BPSK-OFDM was proven to return the best BER performance when only diffracted signals are received in the
channel. In a dispersive environment with multipath interference, system BER can be significantly reduced
by adding CP to the OFDM symbols, however, at the expense of losing 20% of the transmitted signal and
overall data transfer rate. Future work will focus on establishing a full-duplex air-coupled ultrasonic data
communication network for indoor applications by implementing OFDM modulation.
Acknowledgements
This work was supported by Science Foundation Ireland (SFI) Research Frontiers Programme grant number
11/RFP.1/ECE/3119.
References
[1] R. Adler, P. Desmares, J. Spracklen, An ultrasonic remote control for home receivers, Consumer Electronics,
IEEE Transactions on CE-28 (1) (1982) 123–128.
[2] S. Holm, O. Hovind, S. Rostad, R. Holm, Indoors data communications using airborne ultrasound, in:
Acoustics, Speech, and Signal Processing, 2005. Proceedings. (ICASSP ’05). IEEE International Conference
on, Vol. 3, 2005, pp. 957–960.
14
−5 0 5−5
−4
−3
−2
−1
0
1
2
3
4
5
In−Phase(a)
Qua
drat
ure
−5 0 5−5
−4
−3
−2
−1
0
1
2
3
4
5
In−Phase(b)
Qua
drat
ure
Figure 13: Received 16-QAM OFDM signal constellations after equalisation (a) without CP and (b) with CP at a SNR value of
32 dB.
[3] M. Hazas, A. Hopper, Broadband ultrasonic location systems for improved indoor positioning, Mobile
Computing, IEEE Transactions on 5 (5) (2006) 536–547.
[4] C. Li, D. Hutchins, R. Green, Short-range ultrasonic digital communications in air, Ultrasonics, Ferro-
electrics and Frequency Control, IEEE Transactions on 55 (4) (2008) 908–918.
[5] C. Li, D. Hutchins, R. Green, Short-range ultrasonic communications in air using quadrature modulation,
Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on 56 (10) (2009) 2060–2072.
[6] K. Nakamura (Ed.), Ultrasonic Transducers: Materials and Design for Sensors, Actuators and Medical
Applications, Woodhead Publishing Limited, 2012.
[7] S. Weinstein, The history of orthogonal frequency-division multiplexing [history of communications], Com-
munications Magazine, IEEE 47 (11) (2009) 26–35.
[8] W. M. D. Wright, O. Doyle, C. T. Foley, Multi-channel data transfer using air-coupled capacitive ultrasonic
transducers, in: Ultrasonics Symposium, 2006. IEEE, 2006, pp. 1805–1808.
[9] A. Ens, L. Reindl, Multicarrier airborne ultrasound transmission with piezoelectric transducers, Ultrasonics,
Ferroelectrics, and Frequency Control, IEEE Transactions on 62 (5) (2015) 905–914.
[10] J. Proakis, M. Salehi, Digital Communications, McGraw-Hill Higher Education, 2008.
[11] S. Weinstein, P. Ebert, Data transmission by frequency-division multiplexing using the discrete fourier
transform, Communication Technology, IEEE Transactions on 19 (5) (1971) 628–634.
[12] M. Skolnik, Introduction to Radar Systems, Electrical engineering series, McGraw Hill, 2001.
[13] P. Lazik, A. Rowe, Indoor pseudo-ranging of mobile devices using ultrasonic chirps, in: Proceedings of the
10th ACM Conference on Embedded Network Sensor Systems, SenSys ’12, ACM, New York, NY, USA,
2012, pp. 99–112.
[14] D.-C. Chang, Effect and compensation of symbol timing offset in ofdm systems with channel interpolation,
Broadcasting, IEEE Transactions on 54 (4) (2008) 761–770.
15
[15] Senscomp, inc. the senscomp website. (Jan. 2016).
URL http://www.senscomp.com/pdfs/series-600-environmental-grade-sensor.pdf
[16] E. McCune, Practical digital wireless signals, Cambridge University Press, 2010.
[17] M. J. Crocker, Handbook of Acoustics, A Wiley-Interscience Publication, Wiley, 1998.
[18] N. H. F. Thomas D. Rossing, Principles of Vibration and Sound, Springer, 2004.
[19] W. Jiang, W. M. D. Wright, Multichannel ultrasonic data communications in air using range-dependent
modulation schemes, Ultrasonics, Ferroelectrics, and Frequency Control, IEEE Transactions on 63 (1)
(2016) 147–155.
16